[Gvsig_english] Feature request: progress bar when loading big layers

[Cèsar Ordiñana]

El 31/03/11 18:50, Antonio Falciano escribió:
>
>> The question about loading vectorfiles is interesting, I think. As
>> I've got already problems with raster files @ 25MB...(start parameter
>> at 256MB)
>>
>> I've tested some shapefiles (from http://downloads.cloudmade.com)
>> with the gvSIG Mobile src running in Eclipse (OS: Windows 7):
>>
>>      File: france_highways.shp (size = 336277700)
>>      Error: "Error while opening file: Not enough memory" =>  no layer
>>      in ToC, but it is loaded into the map view!??
>>
>>      File: spain_highways.shp (size = 185766396)
>>      Error: same error message
>>
>>      File: germany_natural.shp (size = 152891836)
>>      Error: no problems!
>>
>>      File: germany_highways.shp (size = 810905604)
>>      Error: not enough memory - of course
>>
>>
>> The problems seem to start at about 150MB? Does it depend on the
>> computer or does it depend on java itself? is there any differences
>> using gvsig mobile or gvsig desktop (while loading shape files) ?
> Your test is very interesting and useful for debugging, I think. The
> bottleneck should depend by a combination of these factors, plus the
> particular used driver. For instance, I was not able to load a DXF layer
> of 60 MB on gvSIG Desktop 1.10 today.
>
> Cheers,
> Antonio

Yes, one of the main factors is the driver, some of them load everything 
into memory (ex: DXF), others read everything from disc (ex: SHP).

I think it's very difficult to define a size limit for vectorial layer 
files, because there are a lot of factors involved:

- Driver type.
- Geometry complexity.
- Available alphanumeric data.
- If any operation has been performed in the layer: symbology, labeling, 
ordering, filtering, etc.
...

At least, a better behavior when loading a layer which does not fit in 
memory could be developed, notifying about the error and not loading an 
empty layer. But its only a supposition, some investigation would need 
to be performed to see if it is feasible.

About the progress bar, I agree also. At least a waiting cursor could be 
used.

I think you could create two new feature requests about those two 
problems, so they don't get forgotten.

Regards,

-- 
Cèsar Ordiñana Navarro
gvSIG software architect
DiSiD Technologies (http://www.disid.com)